PALM: Personalized Active Learning for mmWave-Based Activity Recognition
Bibliographic record
Abstract
Human activity recognition (HAR) plays a crucial role in enhancing human safety and well-being, with applications ranging from dietary management to driver monitoring. Millimeter-wave (mmWave) radars have emerged as a promising technology for HAR due to their ability to capture fine-grained activities without the inconvenience associated with wearable sensors or the potential privacy issues posed by vision-based sensors. In this article, we introduce personalized active learning for mmWave (PALM). Built upon our previously proposed dynamic point cloud recognizer (DPR), PALM addresses the challenge of cold start for new users by utilizing uncertainty to selectively query the user about the most informative samples while training a personalized model. Experiments on our food intake activity dataset (FIAD) demonstrate that PALM attains 91.08% accuracy over a two-week active learning period, surpassing the baseline and alternative uncertainty quantification methods. Furthermore, leveraging transfer learning from our driver activity dataset (DAD), PALM achieves a 9.87% higher accuracy and 19.48% improvement in area under the curve (AUC) compared to the baseline model trained from scratch. These results highlight PALM’s effectiveness in personalizing HAR models while minimizing labeling effort, making it suitable for widespread deployment in real-world applications. In addition, we show that DPR outperforms state-of-the-art voxelization-based methods, achieving a 4.10% increase in accuracy while reducing memory consumption by 78.29% and inference time by 69.64%, leading to resource efficiency.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.004 | 0.002 |
| Research integrity | 0.001 | 0.003 |
| Insufficient payload (model declined to judge) | 0.003 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".